When Sara Du was helping businesses build Model Context Protocol (MCP) servers in 2025, she kept fielding the same question from clients: how could AI agents be used directly within tools like Slack? The problem, it turned out, was more than just a technical inconvenience. Shuttling messages between systems, giving agents the right context at the right time, and doing all of this without running up enormous token costs was proving to be a persistent challenge with no clean solution in sight.

As Du dug deeper into the issue, she came to believe the problem was not simply a matter of integration. It was structural. The platforms themselves had been built around assumptions that were becoming obsolete. Slack, Microsoft Teams, and similar tools were designed for human-to-human communication, and AI agents were being bolted on as an afterthought - treated like installable apps rather than genuine participants in workplace conversations.

What Du identified as particularly frustrating was a pattern she began calling "meat proxies" - a term she uses to describe the situation where a human worker ends up acting as a go-between, manually relaying an AI agent's output to the rest of a team. In her view, the human becomes little more than a messenger in a system that should be capable of communicating on its own.

Her proposed alternative was a platform where agents could participate directly in the shared spaces where work actually happens - reading the history behind decisions, asking questions of colleagues, building on the output of other agents, and escalating to a human only when a judgment call is genuinely required.

A Messaging Platform Built Around AI Participation

That vision has now taken the form of a product. Du's startup, Ando, emerged from stealth on Thursday with the launch of a team messaging application built from the ground up to accommodate both human workers and AI agents as first-class participants. The company is positioning Ando as a full replacement for existing workplace messaging platforms - not a plugin or an add-on, but an entirely new environment designed for the way AI-integrated teams actually operate.

In practical terms, Ando gives AI agents their own identities and inboxes within the platform. Agents can join channels, participate in group conversations, and engage in direct messages in much the same way a human team member would. The platform supports live calls that can be transcribed and made accessible to agents in real time. Agents are not limited to responding when tagged - they can browse available channels, choose which ones to join based on relevance, and proactively reach out to human colleagues when they determine that a notification or update is warranted, without needing to wait for a person to prompt them first.

Funding and Investor Backing

To support continued development of the platform, Ando announced on Thursday that it has secured $20 million in combined pre-seed and seed funding. The round includes backing from prominent venture capital firms Accel, Index Ventures, and Emergence Capital. The funds are earmarked for expanding the team and scaling the infrastructure required to handle the volume of AI-driven interactions the platform is designed to support.

A Crowded but Evolving Market

The concept of integrating AI agents into workplace communication tools is not new territory. The major incumbents in the enterprise collaboration space have all moved in this direction to varying degrees. Slack has transformed its native bot functionality into a fully capable AI agent. Microsoft has made substantial investments in embedding Copilot throughout Teams, with deep connections to the broader Office 365 ecosystem. These are well-resourced efforts backed by companies with enormous existing user bases.

Even so, there remains room for challengers. The pace at which AI is reshaping workplace dynamics has created openings for products that take a fundamentally different architectural approach. Earlier this year, Jack Dorsey - co-founder of both Twitter and Block - unveiled a product called Buzz, a messaging application designed to bring humans and AI agents together in a shared environment, though that product is currently oriented more toward developers than general enterprise teams.

Du's argument is that building a platform specifically designed for AI-agent participation from day one carries inherent advantages over legacy platforms attempting to retrofit AI capabilities into software originally built for a different era. The architectural decisions made early in a product's development tend to have long-lasting consequences, and she believes those constraints will limit how far existing platforms can evolve.

A Slow Start and a Turning Point

Du has been candid about the challenges Ando faced in its early days. Initial reactions from people who saw the product were often underwhelming. At its earliest stage, Ando was, by her own admission, simply a less polished version of messaging tools people already used. Many potential users focused on that comparison and did not get far enough into the experience to appreciate what made the agent integration distinctive.

The shift came gradually, as more customers began using the platform for sustained periods and the behavior of agents within the system started to reveal capabilities that went beyond what most people anticipated. One example that Du found particularly striking: an agent independently noticed two separate conversations happening in different channels, both circling the same underlying problem. Without being prompted, the agent created a group chat, brought the relevant participants together, summarized the context from both threads, and proposed a course of action.

That kind of autonomous coordination - identifying redundancy, consolidating effort, and facilitating a decision - illustrated something Du had come to believe more broadly: that AI agents, given the right environment, can take on coordination tasks that humans often struggle to manage at scale, simply because agents can process far greater volumes of information in far less time.

Current Customers and the Road Ahead

Ando is currently working with customers across the software, real estate, and finance sectors, with teams spread across 15 countries. Most of the teams using the platform at this stage are relatively small, which aligns with Du's broader thesis about where AI-driven collaboration tools will have the most immediate impact.

Her view is that AI agents will enable small teams to operate at a scale that would have previously required organizations many times their size. By taking on execution tasks, research, and coordination work, agents free up human workers to concentrate on the kinds of decisions that require judgment, strategic thinking, and the ability to determine what should happen next - areas where human input remains essential.

The fresh capital will go toward hiring and toward the infrastructure costs associated with running a platform where AI agents are continuously active participants rather than occasional tools. Whether Ando can carve out a durable position in a market where well-funded incumbents are moving quickly remains to be seen, but the company is entering the conversation with a clear architectural argument and early evidence that its approach resonates with at least some portion of the teams navigating the transition to AI-integrated work.